Models with low observed uncertainty compared to rv uncertainty

[junpenglao]

Difficult to say without the data and the model, but since:

@NateAM

the variation in my measured test data is order 1e-3 - 1e-4

did you try scaling the measured data?

For example, puting it into a range of 0 - 1 or 0 - 10 would be easier to handle numerically.